AI’s Climate Promise Falters as Industry Lags Behind COP28 Expectations
AI’s climate impact remains largely theoretical, with only a handful of real‑world deployments matching ambitious emission‑reduction forecasts.
S
··2 min readSciTech Editorial Team
Newsroom

As world leaders gather in Antalya for the United Nations climate summit, artificial intelligence (AI) dominates the agenda, yet the technology’s climate case remains largely unproven. The host nation has urged nations to ensure AI is designed, powered, and deployed in ways that bolster climate action, but industry-wide adoption falls far short of the lofty claims made by AI firms.
A peer‑reviewed study from the London School of Economics’ Grantham Research Institute and consultancy Systemiq projected that AI could slash global carbon dioxide emissions by 3.2 to 5.4 billion tonnes per year by 2035 – a reduction comparable to the United States’ entire annual energy‑related output. The bulk of those gains were expected in power generation, meat and dairy, and transportation, with AI‑driven supply‑demand forecasting potentially boosting renewable electricity penetration by up to 20% and averting roughly 1.8 billion tonnes of CO₂ annually. Yet, a systematic review of public disclosures shows that only the UK’s National Grid ESO, in partnership with Open Climate Fix, has deployed a meaningful AI forecasting tool, halving solar‑forecast error and avoiding hundreds of thousands of tonnes of CO₂ each year.
In the meat and dairy arena, the gap between technological readiness and market reality is stark. NotCo’s joint venture with Kraft Heinz leverages its Giuseppe AI platform to match plant‑based molecules to the taste, texture, and aroma of animal products, representing one of the few deep integrations. However, most alternative‑protein firms lack comparable AI deployment, and the sector faces dwindling venture capital – funding fell from a $989 million peak in 2021 to about $83 million last year – hampering broader adoption.
Transportation offers more encouraging signals. CATL, which commands roughly 40% of the global electric‑vehicle battery market, introduced an AI system trained on over 100,000 design cases and 600 TB of test data, cutting material‑screening time by two orders of magnitude. Independent analysis also shows that ride‑hailing giant Didi Chuxing reduced per‑trip emissions by an average of 16% through AI‑optimized routing, illustrating how AI can deliver tangible efficiency gains.
Nevertheless, the climate narrative for AI is complicated by parallel interests from oil and gas firms, which seek AI to boost fossil‑fuel production. Moreover, AI alone cannot accelerate the lengthy certification, permitting, and market‑entry processes that new technologies require. As the summit concludes, policymakers and tech leaders must confront the disparity between AI’s projected climate impact and the modest, uneven progress observed on the ground.



